[wplug] wow

Vanco, Donald VANCOD at PIOS.com
Wed Sep 24 10:04:54 EDT 2003


Lance Tost <mailto:ltost at pobox.com> scribbled on Tuesday, September 23, 2003
9:35 PM:
> I'm a huge fan of Linux.  I've been a user and admin for over 10
> years. But I have to disagree with the statement about Sun.  
	Not arguing anyone's personal experience/opinion, but at one point
in time I was getting a 50% failure rate on Sun hardware.  I had so many
Ultra30's bombing on me that the guys on the support line had one on the
dock ready to ship when I called with a failure.  
	I didn't really have too many issues with the OS (but certainly no
better or worse than IRIX or Linux or HP-UX), and frankly owe what minimal
SysAdmin skilz I possess to Solaris.
	I think my overall bend is the waffling that Sun has shown over the
years on a couple fronts - the support for Solaris x86 (which is OT) and
support for Linux.  Listening to Sun is like watching Sybil
http://www.imdb.com/title/tt0075296/.  I was at LinuxWorld 2 years ago and
listened to McNealy give a keynote in which he RAVED about Linux but slammed
on Red Hat - yet I was able to snap a photo of a "Sun Linux" box booting in
their booth with a Red Hat start-up screen.  I can't seem to find that
particular photo at the moment, but I do have one that, IMHO, shows about
the best use of a Sun server:
http://www.angelfire.com/geek/compugeek/Sun_SERVER.jpg

> RedHat's up2date has bitten me a couple
> of times by overwriting apache configs, and occasionally updating my
> kernel to one that won't boot correctly.  
	I too have had issues with up2date - in my case annihilating Mozilla
and destroying all "bluecurse" menu icons.  As far as kernel updates, well,
that's always had a "caveat emptor" feel to it IMHO.  I have set up2date to
d/l only, and then run:
rpm -Uvh $(ls *.rpm | egrep -v '^(kernel-)')  
....from /var/spool/up2date
	I then run an install (as opposed to "freshen") on the kernels and
will remove the old ones after a few days of stable operation.  I do not
understand why RH does not config up2date to handle kernels in an alternate
method - but then it does, by default, bypass kernel updates.

	It's also worth noting that there's a setting in up2date that allows
you to NOT upgrade packages whose local config files have been changed.
"up2date --configure" is your friend.

	On a related note - past experience has shown that the RPM database
can sometimes get confused about installed kernels (for some reason) -
running this:
rpm -q kernel kernel-smp --queryformat \
"%{NAME}-%{VERSION}-%{RELEASE}.%{ARCH}\n"

..tells me what the RPM database sees and I can then "rpm -e" the entire
string (e.g. kernel-smp-2.4.18-3.i686) otherwise I have to hand delete all
the files.  Attempts to remove "kernel-2.4.18-3 will fail (and I'm sure
that's by design)

> BTW, higher end Sun boxes
> allow you to shutdown individual CPU and/or memory boards, remove
> them, and replace them all while the OS (and any apps) continue to
> run.  I believe if we decided tomorrow that our 4 CPUs weren't enough
> for our SAP system, we could throw in a third board with 2 more CPUs
> and a couple more gig of RAM and dynamically configure it into the
> running OS.  I don't believe Linux can do that (yet).  
	RAM and PCI cards - yes, possible today on appropriate hardware.
CPUs - not yet (but IBM probably has it working in a lab, and for 2 years).

> FWIW, I don't
> feel that "Linux doesn't belong on the server" but I do feel that
> Solaris is still a better choice in some situations. 
	I'd agree - Linux is not a panacea.  I suppose in Sun's defense they
didn't say Linux was just plain bad - but they are saying it's not a server
OS.  Funny, there's an @$$load of benchmarks for Enterprise class
applications (Oracle, Sendmail, Reuters, SAP-R/3, etc) that kind of beg to
differ.......

Don



More information about the wplug mailing list